Our studies using positron emission tomography (PET) showed that drugs which interact with the GABAergic system may be more likely to reduce CMRglc than those which do not. We found that valproic acid depressed CBF in the thalamus to a greater degree than in other regions. This effect may be related to the drug's efficacy against primary generalized seizures. In a double-blind placebo-controlled parallel design trial, we are investigating the effect of gamma-vinyl-GABA (GVG) an experimental antiepileptic drug, on cerebral glucose metabolism (CMRglc), blood flow (CBF), CSF GABA levels, and seizure frequency, in patients with uncontrolled complex partial seizures (CPS). Vigabatrin raises CSF GABA, and this effect may be related to seizure control. We are particularly interested in the possibility of regional reductions in LCMRglc or CBF, which predict efficacy or toxicity, including psychiatric disorders. Viga~batrin causes intramyelinic edema in laboratory animals, which has not been found on human MRI or evoked potential studies. PET may be more sensitive than these modalities to subtle abnormalities. Patients who have persistent seizures on carbamazepine therapy (CBZ) will be entered into the study. After initial evaluation to confirm seizure type, establish therapeutic CBZ levels, and exclude patients with evidence of systemic or progressive neurological disease, a baseline PET scan, lumbar puncture, MRI, evoked potential and neuropsychological test battery will be performed. Patients then will be randomized to placebo or vigabatrin in addition to continuous CBZ therapy. Vigabatrin / placebo will be titrated upwards to 50 mg / kg over 3 weeks. After two months the PET scan and other tests will be repeated. The primary outcome measure of the study will be the drug effect on CBF and LCMRglc.